git.ozlabs.org Git - ccan/blob - ccan/tdb2/hash.c

   1  /*
   2    Trivial Database 2: hash handling
   3    Copyright (C) Rusty Russell 2010
   4
   5    This library is free software; you can redistribute it and/or
   6    modify it under the terms of the GNU Lesser General Public
   7    License as published by the Free Software Foundation; either
   8    version 3 of the License, or (at your option) any later version.
   9
  10    This library is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13    Lesser General Public License for more details.
  14
  15    You should have received a copy of the GNU Lesser General Public
  16    License along with this library; if not, see <http://www.gnu.org/licenses/>.
  17 */
  18 #include "private.h"
  19 #include <assert.h>
  20 #include <ccan/hash/hash.h>
  21
  22 static uint64_t jenkins_hash(const void *key, size_t length, uint64_t seed,
  23                              void *arg)
  24 {
  25         uint64_t ret;
  26         /* hash64_stable assumes lower bits are more important; they are a
  27          * slightly better hash.  We use the upper bits first, so swap them. */
  28         ret = hash64_stable((const unsigned char *)key, length, seed);
  29         return (ret >> 32) | (ret << 32);
  30 }
  31
  32 void tdb_hash_init(struct tdb_context *tdb)
  33 {
  34         tdb->khash = jenkins_hash;
  35         tdb->hash_priv = NULL;
  36 }
  37
  38 uint64_t tdb_hash(struct tdb_context *tdb, const void *ptr, size_t len)
  39 {
  40         return tdb->khash(ptr, len, tdb->hash_seed, tdb->hash_priv);
  41 }
  42
  43 uint64_t hash_record(struct tdb_context *tdb, tdb_off_t off)
  44 {
  45         const struct tdb_used_record *r;
  46         const void *key;
  47         uint64_t klen, hash;
  48
  49         r = tdb_access_read(tdb, off, sizeof(*r), true);
  50         if (!r)
  51                 /* FIXME */
  52                 return 0;
  53
  54         klen = rec_key_length(r);
  55         tdb_access_release(tdb, r);
  56
  57         key = tdb_access_read(tdb, off + sizeof(*r), klen, false);
  58         if (!key)
  59                 return 0;
  60
  61         hash = tdb_hash(tdb, key, klen);
  62         tdb_access_release(tdb, key);
  63         return hash;
  64 }
  65
  66 /* Get bits from a value. */
  67 static uint32_t bits_from(uint64_t val, unsigned start, unsigned num)
  68 {
  69         assert(num <= 32);
  70         return (val >> start) & ((1U << num) - 1);
  71 }
  72
  73 /* We take bits from the top: that way we can lock whole sections of the hash
  74  * by using lock ranges. */
  75 static uint32_t use_bits(struct hash_info *h, unsigned num)
  76 {
  77         h->hash_used += num;
  78         return bits_from(h->h, 64 - h->hash_used, num);
  79 }
  80
  81 static bool key_matches(struct tdb_context *tdb,
  82                         const struct tdb_used_record *rec,
  83                         tdb_off_t off,
  84                         const struct tdb_data *key)
  85 {
  86         bool ret = false;
  87         const char *rkey;
  88
  89         if (rec_key_length(rec) != key->dsize) {
  90                 add_stat(tdb, compare_wrong_keylen, 1);
  91                 return ret;
  92         }
  93
  94         rkey = tdb_access_read(tdb, off + sizeof(*rec), key->dsize, false);
  95         if (!rkey)
  96                 return ret;
  97         if (memcmp(rkey, key->dptr, key->dsize) == 0)
  98                 ret = true;
  99         else
 100                 add_stat(tdb, compare_wrong_keycmp, 1);
 101         tdb_access_release(tdb, rkey);
 102         return ret;
 103 }
 104
 105 /* Does entry match? */
 106 static bool match(struct tdb_context *tdb,
 107                   struct hash_info *h,
 108                   const struct tdb_data *key,
 109                   tdb_off_t val,
 110                   struct tdb_used_record *rec)
 111 {
 112         tdb_off_t off;
 113
 114         add_stat(tdb, compares, 1);
 115         /* Desired bucket must match. */
 116         if (h->home_bucket != (val & TDB_OFF_HASH_GROUP_MASK)) {
 117                 add_stat(tdb, compare_wrong_bucket, 1);
 118                 return false;
 119         }
 120
 121         /* Top bits of offset == next bits of hash. */
 122         if (bits_from(val, TDB_OFF_HASH_EXTRA_BIT, TDB_OFF_UPPER_STEAL_EXTRA)
 123             != bits_from(h->h, 64 - h->hash_used - TDB_OFF_UPPER_STEAL_EXTRA,
 124                     TDB_OFF_UPPER_STEAL_EXTRA)) {
 125                 add_stat(tdb, compare_wrong_offsetbits, 1);
 126                 return false;
 127         }
 128
 129         off = val & TDB_OFF_MASK;
 130         if (tdb_read_convert(tdb, off, rec, sizeof(*rec)) == -1)
 131                 return false;
 132
 133         if ((h->h & ((1 << 11)-1)) != rec_hash(rec)) {
 134                 add_stat(tdb, compare_wrong_rechash, 1);
 135                 return false;
 136         }
 137
 138         return key_matches(tdb, rec, off, key);
 139 }
 140
 141 static tdb_off_t hbucket_off(tdb_off_t group_start, unsigned bucket)
 142 {
 143         return group_start
 144                 + (bucket % (1 << TDB_HASH_GROUP_BITS)) * sizeof(tdb_off_t);
 145 }
 146
 147 bool is_subhash(tdb_off_t val)
 148 {
 149         return (val >> TDB_OFF_UPPER_STEAL_SUBHASH_BIT) & 1;
 150 }
 151
 152 /* FIXME: Guess the depth, don't over-lock! */
 153 static tdb_off_t hlock_range(tdb_off_t group, tdb_off_t *size)
 154 {
 155         *size = 1ULL << (64 - (TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS));
 156         return group << (64 - (TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS));
 157 }
 158
 159 static tdb_off_t COLD find_in_chain(struct tdb_context *tdb,
 160                                     struct tdb_data key,
 161                                     tdb_off_t chain,
 162                                     struct hash_info *h,
 163                                     struct tdb_used_record *rec,
 164                                     struct traverse_info *tinfo)
 165 {
 166         tdb_off_t off, next;
 167
 168         /* In case nothing is free, we set these to zero. */
 169         h->home_bucket = h->found_bucket = 0;
 170
 171         for (off = chain; off; off = next) {
 172                 unsigned int i;
 173
 174                 h->group_start = off;
 175                 if (tdb_read_convert(tdb, off, h->group, sizeof(h->group)))
 176                         return TDB_OFF_ERR;
 177
 178                 for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 179                         tdb_off_t recoff;
 180                         if (!h->group[i]) {
 181                                 /* Remember this empty bucket. */
 182                                 h->home_bucket = h->found_bucket = i;
 183                                 continue;
 184                         }
 185
 186                         /* We can insert extra bits via add_to_hash
 187                          * empty bucket logic. */
 188                         recoff = h->group[i] & TDB_OFF_MASK;
 189                         if (tdb_read_convert(tdb, recoff, rec, sizeof(*rec)))
 190                                 return TDB_OFF_ERR;
 191
 192                         if (key_matches(tdb, rec, recoff, &key)) {
 193                                 h->home_bucket = h->found_bucket = i;
 194
 195                                 if (tinfo) {
 196                                         tinfo->levels[tinfo->num_levels]
 197                                                 .hashtable = off;
 198                                         tinfo->levels[tinfo->num_levels]
 199                                                 .total_buckets
 200                                                 = 1 << TDB_HASH_GROUP_BITS;
 201                                         tinfo->levels[tinfo->num_levels].entry
 202                                                 = i;
 203                                         tinfo->num_levels++;
 204                                 }
 205                                 return recoff;
 206                         }
 207                 }
 208                 next = tdb_read_off(tdb, off
 209                                     + offsetof(struct tdb_chain, next));
 210                 if (next == TDB_OFF_ERR)
 211                         return TDB_OFF_ERR;
 212                 if (next)
 213                         next += sizeof(struct tdb_used_record);
 214         }
 215         return 0;
 216 }
 217
 218 /* This is the core routine which searches the hashtable for an entry.
 219  * On error, no locks are held and TDB_OFF_ERR is returned.
 220  * Otherwise, hinfo is filled in (and the optional tinfo).
 221  * If not found, the return value is 0.
 222  * If found, the return value is the offset, and *rec is the record. */
 223 tdb_off_t find_and_lock(struct tdb_context *tdb,
 224                         struct tdb_data key,
 225                         int ltype,
 226                         struct hash_info *h,
 227                         struct tdb_used_record *rec,
 228                         struct traverse_info *tinfo)
 229 {
 230         uint32_t i, group;
 231         tdb_off_t hashtable;
 232         enum TDB_ERROR ecode;
 233
 234         h->h = tdb_hash(tdb, key.dptr, key.dsize);
 235         h->hash_used = 0;
 236         group = use_bits(h, TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS);
 237         h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 238
 239         h->hlock_start = hlock_range(group, &h->hlock_range);
 240         ecode = tdb_lock_hashes(tdb, h->hlock_start, h->hlock_range, ltype,
 241                                 TDB_LOCK_WAIT);
 242         if (ecode != TDB_SUCCESS) {
 243                 tdb->ecode = ecode;
 244                 return TDB_OFF_ERR;
 245         }
 246
 247         hashtable = offsetof(struct tdb_header, hashtable);
 248         if (tinfo) {
 249                 tinfo->toplevel_group = group;
 250                 tinfo->num_levels = 1;
 251                 tinfo->levels[0].entry = 0;
 252                 tinfo->levels[0].hashtable = hashtable
 253                         + (group << TDB_HASH_GROUP_BITS) * sizeof(tdb_off_t);
 254                 tinfo->levels[0].total_buckets = 1 << TDB_HASH_GROUP_BITS;
 255         }
 256
 257         while (h->hash_used <= 64) {
 258                 /* Read in the hash group. */
 259                 h->group_start = hashtable
 260                         + group * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 261
 262                 if (tdb_read_convert(tdb, h->group_start, &h->group,
 263                                      sizeof(h->group)) == -1)
 264                         goto fail;
 265
 266                 /* Pointer to another hash table?  Go down... */
 267                 if (is_subhash(h->group[h->home_bucket])) {
 268                         hashtable = (h->group[h->home_bucket] & TDB_OFF_MASK)
 269                                 + sizeof(struct tdb_used_record);
 270                         if (tinfo) {
 271                                 /* When we come back, use *next* bucket */
 272                                 tinfo->levels[tinfo->num_levels-1].entry
 273                                         += h->home_bucket + 1;
 274                         }
 275                         group = use_bits(h, TDB_SUBLEVEL_HASH_BITS
 276                                          - TDB_HASH_GROUP_BITS);
 277                         h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 278                         if (tinfo) {
 279                                 tinfo->levels[tinfo->num_levels].hashtable
 280                                         = hashtable;
 281                                 tinfo->levels[tinfo->num_levels].total_buckets
 282                                         = 1 << TDB_SUBLEVEL_HASH_BITS;
 283                                 tinfo->levels[tinfo->num_levels].entry
 284                                         = group << TDB_HASH_GROUP_BITS;
 285                                 tinfo->num_levels++;
 286                         }
 287                         continue;
 288                 }
 289
 290                 /* It's in this group: search (until 0 or all searched) */
 291                 for (i = 0, h->found_bucket = h->home_bucket;
 292                      i < (1 << TDB_HASH_GROUP_BITS);
 293                      i++, h->found_bucket = ((h->found_bucket+1)
 294                                              % (1 << TDB_HASH_GROUP_BITS))) {
 295                         if (is_subhash(h->group[h->found_bucket]))
 296                                 continue;
 297
 298                         if (!h->group[h->found_bucket])
 299                                 break;
 300
 301                         if (match(tdb, h, &key, h->group[h->found_bucket],
 302                                   rec)) {
 303                                 if (tinfo) {
 304                                         tinfo->levels[tinfo->num_levels-1].entry
 305                                                 += h->found_bucket;
 306                                 }
 307                                 return h->group[h->found_bucket] & TDB_OFF_MASK;
 308                         }
 309                 }
 310                 /* Didn't find it: h indicates where it would go. */
 311                 return 0;
 312         }
 313
 314         return find_in_chain(tdb, key, hashtable, h, rec, tinfo);
 315
 316 fail:
 317         tdb_unlock_hashes(tdb, h->hlock_start, h->hlock_range, ltype);
 318         return TDB_OFF_ERR;
 319 }
 320
 321 /* I wrote a simple test, expanding a hash to 2GB, for the following
 322  * cases:
 323  * 1) Expanding all the buckets at once,
 324  * 2) Expanding the bucket we wanted to place the new entry into.
 325  * 3) Expanding the most-populated bucket,
 326  *
 327  * I measured the worst/average/best density during this process.
 328  * 1) 3%/16%/30%
 329  * 2) 4%/20%/38%
 330  * 3) 6%/22%/41%
 331  *
 332  * So we figure out the busiest bucket for the moment.
 333  */
 334 static unsigned fullest_bucket(struct tdb_context *tdb,
 335                                const tdb_off_t *group,
 336                                unsigned new_bucket)
 337 {
 338         unsigned counts[1 << TDB_HASH_GROUP_BITS] = { 0 };
 339         unsigned int i, best_bucket;
 340
 341         /* Count the new entry. */
 342         counts[new_bucket]++;
 343         best_bucket = new_bucket;
 344
 345         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 346                 unsigned this_bucket;
 347
 348                 if (is_subhash(group[i]))
 349                         continue;
 350                 this_bucket = group[i] & TDB_OFF_HASH_GROUP_MASK;
 351                 if (++counts[this_bucket] > counts[best_bucket])
 352                         best_bucket = this_bucket;
 353         }
 354
 355         return best_bucket;
 356 }
 357
 358 static bool put_into_group(tdb_off_t *group,
 359                            unsigned bucket, tdb_off_t encoded)
 360 {
 361         unsigned int i;
 362
 363         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 364                 unsigned b = (bucket + i) % (1 << TDB_HASH_GROUP_BITS);
 365
 366                 if (group[b] == 0) {
 367                         group[b] = encoded;
 368                         return true;
 369                 }
 370         }
 371         return false;
 372 }
 373
 374 static void force_into_group(tdb_off_t *group,
 375                              unsigned bucket, tdb_off_t encoded)
 376 {
 377         if (!put_into_group(group, bucket, encoded))
 378                 abort();
 379 }
 380
 381 static tdb_off_t encode_offset(tdb_off_t new_off, struct hash_info *h)
 382 {
 383         return h->home_bucket
 384                 | new_off
 385                 | ((uint64_t)bits_from(h->h,
 386                                   64 - h->hash_used - TDB_OFF_UPPER_STEAL_EXTRA,
 387                                   TDB_OFF_UPPER_STEAL_EXTRA)
 388                    << TDB_OFF_HASH_EXTRA_BIT);
 389 }
 390
 391 /* Simply overwrite the hash entry we found before. */
 392 int replace_in_hash(struct tdb_context *tdb,
 393                     struct hash_info *h,
 394                     tdb_off_t new_off)
 395 {
 396         return tdb_write_off(tdb, hbucket_off(h->group_start, h->found_bucket),
 397                              encode_offset(new_off, h));
 398 }
 399
 400 /* We slot in anywhere that's empty in the chain. */
 401 static int COLD add_to_chain(struct tdb_context *tdb,
 402                              tdb_off_t subhash,
 403                              tdb_off_t new_off)
 404 {
 405         size_t entry = tdb_find_zero_off(tdb, subhash, 1<<TDB_HASH_GROUP_BITS);
 406
 407         if (entry == 1 << TDB_HASH_GROUP_BITS) {
 408                 tdb_off_t next;
 409
 410                 next = tdb_read_off(tdb, subhash
 411                                     + offsetof(struct tdb_chain, next));
 412                 if (next == TDB_OFF_ERR)
 413                         return -1;
 414
 415                 if (!next) {
 416                         next = alloc(tdb, 0, sizeof(struct tdb_chain), 0,
 417                                      TDB_CHAIN_MAGIC, false);
 418                         if (next == TDB_OFF_ERR)
 419                                 return -1;
 420                         if (zero_out(tdb, next+sizeof(struct tdb_used_record),
 421                                      sizeof(struct tdb_chain)))
 422                                 return -1;
 423                         if (tdb_write_off(tdb, subhash
 424                                           + offsetof(struct tdb_chain, next),
 425                                           next) != 0)
 426                                 return -1;
 427                 }
 428                 return add_to_chain(tdb, next, new_off);
 429         }
 430
 431         return tdb_write_off(tdb, subhash + entry * sizeof(tdb_off_t),
 432                              new_off);
 433 }
 434
 435 /* Add into a newly created subhash. */
 436 static int add_to_subhash(struct tdb_context *tdb, tdb_off_t subhash,
 437                           unsigned hash_used, tdb_off_t val)
 438 {
 439         tdb_off_t off = (val & TDB_OFF_MASK), *group;
 440         struct hash_info h;
 441         unsigned int gnum;
 442
 443         h.hash_used = hash_used;
 444
 445         if (hash_used + TDB_SUBLEVEL_HASH_BITS > 64)
 446                 return add_to_chain(tdb, subhash, off);
 447
 448         h.h = hash_record(tdb, off);
 449         gnum = use_bits(&h, TDB_SUBLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS);
 450         h.group_start = subhash
 451                 + gnum * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 452         h.home_bucket = use_bits(&h, TDB_HASH_GROUP_BITS);
 453
 454         group = tdb_access_write(tdb, h.group_start,
 455                                  sizeof(*group) << TDB_HASH_GROUP_BITS, true);
 456         if (!group)
 457                 return -1;
 458         force_into_group(group, h.home_bucket, encode_offset(off, &h));
 459         return tdb_access_commit(tdb, group);
 460 }
 461
 462 static int expand_group(struct tdb_context *tdb, struct hash_info *h)
 463 {
 464         unsigned bucket, num_vals, i, magic;
 465         size_t subsize;
 466         tdb_off_t subhash;
 467         tdb_off_t vals[1 << TDB_HASH_GROUP_BITS];
 468
 469         /* Attach new empty subhash under fullest bucket. */
 470         bucket = fullest_bucket(tdb, h->group, h->home_bucket);
 471
 472         if (h->hash_used == 64) {
 473                 add_stat(tdb, alloc_chain, 1);
 474                 subsize = sizeof(struct tdb_chain);
 475                 magic = TDB_CHAIN_MAGIC;
 476         } else {
 477                 add_stat(tdb, alloc_subhash, 1);
 478                 subsize = (sizeof(tdb_off_t) << TDB_SUBLEVEL_HASH_BITS);
 479                 magic = TDB_HTABLE_MAGIC;
 480         }
 481
 482         subhash = alloc(tdb, 0, subsize, 0, magic, false);
 483         if (subhash == TDB_OFF_ERR)
 484                 return -1;
 485
 486         if (zero_out(tdb, subhash + sizeof(struct tdb_used_record), subsize))
 487                 return -1;
 488
 489         /* Remove any which are destined for bucket or are in wrong place. */
 490         num_vals = 0;
 491         for (i = 0; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 492                 unsigned home_bucket = h->group[i] & TDB_OFF_HASH_GROUP_MASK;
 493                 if (!h->group[i] || is_subhash(h->group[i]))
 494                         continue;
 495                 if (home_bucket == bucket || home_bucket != i) {
 496                         vals[num_vals++] = h->group[i];
 497                         h->group[i] = 0;
 498                 }
 499         }
 500         /* FIXME: This assert is valid, but we do this during unit test :( */
 501         /* assert(num_vals); */
 502
 503         /* Overwrite expanded bucket with subhash pointer. */
 504         h->group[bucket] = subhash | (1ULL << TDB_OFF_UPPER_STEAL_SUBHASH_BIT);
 505
 506         /* Point to actual contents of record. */
 507         subhash += sizeof(struct tdb_used_record);
 508
 509         /* Put values back. */
 510         for (i = 0; i < num_vals; i++) {
 511                 unsigned this_bucket = vals[i] & TDB_OFF_HASH_GROUP_MASK;
 512
 513                 if (this_bucket == bucket) {
 514                         if (add_to_subhash(tdb, subhash, h->hash_used, vals[i]))
 515                                 return -1;
 516                 } else {
 517                         /* There should be room to put this back. */
 518                         force_into_group(h->group, this_bucket, vals[i]);
 519                 }
 520         }
 521         return 0;
 522 }
 523
 524 int delete_from_hash(struct tdb_context *tdb, struct hash_info *h)
 525 {
 526         unsigned int i, num_movers = 0;
 527         tdb_off_t movers[1 << TDB_HASH_GROUP_BITS];
 528
 529         h->group[h->found_bucket] = 0;
 530         for (i = 1; i < (1 << TDB_HASH_GROUP_BITS); i++) {
 531                 unsigned this_bucket;
 532
 533                 this_bucket = (h->found_bucket+i) % (1 << TDB_HASH_GROUP_BITS);
 534                 /* Empty bucket?  We're done. */
 535                 if (!h->group[this_bucket])
 536                         break;
 537
 538                 /* Ignore subhashes. */
 539                 if (is_subhash(h->group[this_bucket]))
 540                         continue;
 541
 542                 /* If this one is not happy where it is, we'll move it. */
 543                 if ((h->group[this_bucket] & TDB_OFF_HASH_GROUP_MASK)
 544                     != this_bucket) {
 545                         movers[num_movers++] = h->group[this_bucket];
 546                         h->group[this_bucket] = 0;
 547                 }
 548         }
 549
 550         /* Put back the ones we erased. */
 551         for (i = 0; i < num_movers; i++) {
 552                 force_into_group(h->group, movers[i] & TDB_OFF_HASH_GROUP_MASK,
 553                                  movers[i]);
 554         }
 555
 556         /* Now we write back the hash group */
 557         return tdb_write_convert(tdb, h->group_start,
 558                                  h->group, sizeof(h->group));
 559 }
 560
 561 int add_to_hash(struct tdb_context *tdb, struct hash_info *h, tdb_off_t new_off)
 562 {
 563         /* We hit an empty bucket during search?  That's where it goes. */
 564         if (!h->group[h->found_bucket]) {
 565                 h->group[h->found_bucket] = encode_offset(new_off, h);
 566                 /* Write back the modified group. */
 567                 return tdb_write_convert(tdb, h->group_start,
 568                                          h->group, sizeof(h->group));
 569         }
 570
 571         if (h->hash_used > 64)
 572                 return add_to_chain(tdb, h->group_start, new_off);
 573
 574         /* We're full.  Expand. */
 575         if (expand_group(tdb, h) == -1)
 576                 return -1;
 577
 578         if (is_subhash(h->group[h->home_bucket])) {
 579                 /* We were expanded! */
 580                 tdb_off_t hashtable;
 581                 unsigned int gnum;
 582
 583                 /* Write back the modified group. */
 584                 if (tdb_write_convert(tdb, h->group_start, h->group,
 585                                       sizeof(h->group)))
 586                         return -1;
 587
 588                 /* Move hashinfo down a level. */
 589                 hashtable = (h->group[h->home_bucket] & TDB_OFF_MASK)
 590                         + sizeof(struct tdb_used_record);
 591                 gnum = use_bits(h,TDB_SUBLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS);
 592                 h->home_bucket = use_bits(h, TDB_HASH_GROUP_BITS);
 593                 h->group_start = hashtable
 594                         + gnum * (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 595                 if (tdb_read_convert(tdb, h->group_start, &h->group,
 596                                      sizeof(h->group)) == -1)
 597                         return -1;
 598         }
 599
 600         /* Expanding the group must have made room if it didn't choose this
 601          * bucket. */
 602         if (put_into_group(h->group, h->home_bucket, encode_offset(new_off, h)))
 603                 return tdb_write_convert(tdb, h->group_start,
 604                                          h->group, sizeof(h->group));
 605
 606         /* This can happen if all hashes in group (and us) dropped into same
 607          * group in subhash. */
 608         return add_to_hash(tdb, h, new_off);
 609 }
 610
 611 /* Traverse support: returns offset of record, or 0 or TDB_OFF_ERR. */
 612 static tdb_off_t iterate_hash(struct tdb_context *tdb,
 613                               struct traverse_info *tinfo)
 614 {
 615         tdb_off_t off, val;
 616         unsigned int i;
 617         struct traverse_level *tlevel;
 618
 619         tlevel = &tinfo->levels[tinfo->num_levels-1];
 620
 621 again:
 622         for (i = tdb_find_nonzero_off(tdb, tlevel->hashtable,
 623                                       tlevel->entry, tlevel->total_buckets);
 624              i != tlevel->total_buckets;
 625              i = tdb_find_nonzero_off(tdb, tlevel->hashtable,
 626                                       i+1, tlevel->total_buckets)) {
 627                 val = tdb_read_off(tdb, tlevel->hashtable+sizeof(tdb_off_t)*i);
 628                 if (unlikely(val == TDB_OFF_ERR))
 629                         return TDB_OFF_ERR;
 630
 631                 off = val & TDB_OFF_MASK;
 632
 633                 /* This makes the delete-all-in-traverse case work
 634                  * (and simplifies our logic a little). */
 635                 if (off == tinfo->prev)
 636                         continue;
 637
 638                 tlevel->entry = i;
 639
 640                 if (!is_subhash(val)) {
 641                         /* Found one. */
 642                         tinfo->prev = off;
 643                         return off;
 644                 }
 645
 646                 /* When we come back, we want the next one */
 647                 tlevel->entry++;
 648                 tinfo->num_levels++;
 649                 tlevel++;
 650                 tlevel->hashtable = off + sizeof(struct tdb_used_record);
 651                 tlevel->entry = 0;
 652                 /* Next level is a chain? */
 653                 if (unlikely(tinfo->num_levels == TDB_MAX_LEVELS + 1))
 654                         tlevel->total_buckets = (1 << TDB_HASH_GROUP_BITS);
 655                 else
 656                         tlevel->total_buckets = (1 << TDB_SUBLEVEL_HASH_BITS);
 657                 goto again;
 658         }
 659
 660         /* Nothing there? */
 661         if (tinfo->num_levels == 1)
 662                 return 0;
 663
 664         /* Handle chained entries. */
 665         if (unlikely(tinfo->num_levels == TDB_MAX_LEVELS + 1)) {
 666                 tlevel->hashtable = tdb_read_off(tdb, tlevel->hashtable
 667                                                  + offsetof(struct tdb_chain,
 668                                                             next));
 669                 if (tlevel->hashtable == TDB_OFF_ERR)
 670                         return TDB_OFF_ERR;
 671                 if (tlevel->hashtable) {
 672                         tlevel->hashtable += sizeof(struct tdb_used_record);
 673                         tlevel->entry = 0;
 674                         goto again;
 675                 }
 676         }
 677
 678         /* Go back up and keep searching. */
 679         tinfo->num_levels--;
 680         tlevel--;
 681         goto again;
 682 }
 683
 684 /* Return 1 if we find something, 0 if not, -1 on error. */
 685 int next_in_hash(struct tdb_context *tdb,
 686                  struct traverse_info *tinfo,
 687                  TDB_DATA *kbuf, size_t *dlen)
 688 {
 689         const unsigned group_bits = TDB_TOPLEVEL_HASH_BITS-TDB_HASH_GROUP_BITS;
 690         tdb_off_t hl_start, hl_range, off;
 691         enum TDB_ERROR ecode;
 692
 693         while (tinfo->toplevel_group < (1 << group_bits)) {
 694                 hl_start = (tdb_off_t)tinfo->toplevel_group
 695                         << (64 - group_bits);
 696                 hl_range = 1ULL << group_bits;
 697                 ecode = tdb_lock_hashes(tdb, hl_start, hl_range, F_RDLCK,
 698                                         TDB_LOCK_WAIT);
 699                 if (ecode != TDB_SUCCESS) {
 700                         tdb->ecode = ecode;
 701                         return -1;
 702                 }
 703
 704                 off = iterate_hash(tdb, tinfo);
 705                 if (off) {
 706                         struct tdb_used_record rec;
 707
 708                         if (tdb_read_convert(tdb, off, &rec, sizeof(rec))) {
 709                                 tdb_unlock_hashes(tdb,
 710                                                   hl_start, hl_range, F_RDLCK);
 711                                 return -1;
 712                         }
 713                         if (rec_magic(&rec) != TDB_USED_MAGIC) {
 714                                 tdb_logerr(tdb, TDB_ERR_CORRUPT,
 715                                            TDB_LOG_ERROR,
 716                                            "next_in_hash:"
 717                                            " corrupt record at %llu",
 718                                            (long long)off);
 719                                 return -1;
 720                         }
 721
 722                         kbuf->dsize = rec_key_length(&rec);
 723
 724                         /* They want data as well? */
 725                         if (dlen) {
 726                                 *dlen = rec_data_length(&rec);
 727                                 kbuf->dptr = tdb_alloc_read(tdb,
 728                                                             off + sizeof(rec),
 729                                                             kbuf->dsize
 730                                                             + *dlen);
 731                         } else {
 732                                 kbuf->dptr = tdb_alloc_read(tdb,
 733                                                             off + sizeof(rec),
 734                                                             kbuf->dsize);
 735                         }
 736                         tdb_unlock_hashes(tdb, hl_start, hl_range, F_RDLCK);
 737                         return kbuf->dptr ? 1 : -1;
 738                 }
 739
 740                 tdb_unlock_hashes(tdb, hl_start, hl_range, F_RDLCK);
 741
 742                 tinfo->toplevel_group++;
 743                 tinfo->levels[0].hashtable
 744                         += (sizeof(tdb_off_t) << TDB_HASH_GROUP_BITS);
 745                 tinfo->levels[0].entry = 0;
 746         }
 747         return 0;
 748 }
 749
 750 /* Return 1 if we find something, 0 if not, -1 on error. */
 751 int first_in_hash(struct tdb_context *tdb,
 752                   struct traverse_info *tinfo,
 753                   TDB_DATA *kbuf, size_t *dlen)
 754 {
 755         tinfo->prev = 0;
 756         tinfo->toplevel_group = 0;
 757         tinfo->num_levels = 1;
 758         tinfo->levels[0].hashtable = offsetof(struct tdb_header, hashtable);
 759         tinfo->levels[0].entry = 0;
 760         tinfo->levels[0].total_buckets = (1 << TDB_HASH_GROUP_BITS);
 761
 762         return next_in_hash(tdb, tinfo, kbuf, dlen);
 763 }
 764
 765 /* Even if the entry isn't in this hash bucket, you'd have to lock this
 766  * bucket to find it. */
 767 static int chainlock(struct tdb_context *tdb, const TDB_DATA *key,
 768                      int ltype, enum tdb_lock_flags waitflag,
 769                      const char *func)
 770 {
 771         enum TDB_ERROR ecode;
 772         uint64_t h = tdb_hash(tdb, key->dptr, key->dsize);
 773         tdb_off_t lockstart, locksize;
 774         unsigned int group, gbits;
 775
 776         gbits = TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS;
 777         group = bits_from(h, 64 - gbits, gbits);
 778
 779         lockstart = hlock_range(group, &locksize);
 780
 781         ecode = tdb_lock_hashes(tdb, lockstart, locksize, ltype, waitflag);
 782         tdb_trace_1rec(tdb, func, *key);
 783         if (ecode != TDB_SUCCESS) {
 784                 tdb->ecode = ecode;
 785                 return -1;
 786         }
 787         return 0;
 788 }
 789
 790 /* lock/unlock one hash chain. This is meant to be used to reduce
 791    contention - it cannot guarantee how many records will be locked */
 792 int tdb_chainlock(struct tdb_context *tdb, TDB_DATA key)
 793 {
 794         return chainlock(tdb, &key, F_WRLCK, TDB_LOCK_WAIT, "tdb_chainlock");
 795 }
 796
 797 int tdb_chainunlock(struct tdb_context *tdb, TDB_DATA key)
 798 {
 799         uint64_t h = tdb_hash(tdb, key.dptr, key.dsize);
 800         tdb_off_t lockstart, locksize;
 801         unsigned int group, gbits;
 802         enum TDB_ERROR ecode;
 803
 804         gbits = TDB_TOPLEVEL_HASH_BITS - TDB_HASH_GROUP_BITS;
 805         group = bits_from(h, 64 - gbits, gbits);
 806
 807         lockstart = hlock_range(group, &locksize);
 808
 809         tdb_trace_1rec(tdb, "tdb_chainunlock", key);
 810         ecode = tdb_unlock_hashes(tdb, lockstart, locksize, F_WRLCK);
 811         if (ecode != TDB_SUCCESS) {
 812                 tdb->ecode = ecode;
 813                 return -1;
 814         }
 815         return 0;
 816 }